Loyalty programs for independent restaurants.

Start with one number: on a roughly 500-customer base, the best loyalty model in my own operator practice returned about $6,500 a year in kept margin — and the worst one still worth running returned about $750. Same restaurant, same year, same customers. The two programs sat one menu of choices apart, and the dollars between them ran almost nine to one. These are illustrative figures from a single restaurant, not industry benchmarks — but the order of magnitude is the whole argument, and the order of magnitude barely moves.

I ran two of these programs for over a year on the same restaurant, then dropped both and replaced them with a third. The numbers below are mine, not industry benchmarks. Your conversion rates and average check will differ. The order of magnitude is what carries the argument: the gap between the best and worst model isn’t 20%, it’s 4x.

Four models compete for the same setup hour. The dollars between the best and the worst worth running run almost nine to one — which means the model you pick matters more than how well you run it.

The four models in play

Every loyalty program a US independent restaurant runs fits one of four shapes. The implementations differ; the underlying economics don’t. Walk through them in order of operator cost-to-implement.

Cost-to-implement is the wrong axis to rank these on, because the cheapest model and the most expensive one are both live answers depending on the channel you already run. The axis that actually decides it is annual kept margin — the dollars the program returns after its own cost. On a roughly 500-customer base, in my own operator practice, the spread is the whole argument: the best model returns close to nine times the worst one still worth running. These are illustrative figures from one restaurant, not industry benchmarks.

Why the CRM-direct path wins by 2x

The three competitor models — punch cards, POS loyalty, standalone loyalty platforms — share a structural problem. They all treat every customer the same. The frequent guest who already comes in every week gets the same 10%-off-after-10-visits offer as the once-a-quarter customer who needs an entirely different intervention. The reward is undifferentiated, so the spend is undifferentiated, so the margin lift is mediocre across the board.

The CRM-direct path lets you segment. A weekly regular gets early access to the chef’s upcoming tasting menu and zero discounts — they don’t need a discount, they need to feel known. A lapsed customer who hasn’t visited in 60 days gets a specific $15-off-your-next-visit email at day 58 (right before they’d normally lapse to 90+ days). A first-time direct buyer gets a thank-you with their next-time-save offer. Three customers, three different interventions, three different margin profiles. That’s where the 2x comes from.

Why the punch card still beats the standalone platform on ROI

This is the one that surprises most operators. The punch card costs essentially nothing — ten dollars a year for printing — and produces a real, if small, behavioral lift in repeat visits. Even modest absolute returns make the percentage return enormous. The standalone loyalty platform produces somewhat larger absolute returns, but its cost structure is contact-for-quote (Fivestars and Como do not publish open pricing) and historically lands well above the POS-integrated tier — high enough relative to the behavioral lift that the percentage return is materially worse than either the punch card or POS loyalty. If $200/month is on the table for customer marketing, it’s rarely best-spent on the standalone platform; the same $200 goes further almost anywhere else.

The integrated POS loyalty sits between the two: the cost is low enough that the modest behavioral lift still pays back, but only just — and the data is owned by the POS vendor, so it doesn’t come with you when you switch processors.

What this means for your setup

If you have a direct channel that’s already converting: run the CRM-direct path. Klaviyo for the segmentation; your own Stripe/Toast/Square checkout for the order data; a manually-curated regulars list for the highest-value 50 customers; automated re-engagement at day 58. Vendor-published monthly cost runs $20–$120 depending on email send volume and platform tier. Margin lift in year one varies widely by restaurant; operators with healthy direct channels report the strongest returns of any of the four models in this comparison.

If you don’t have a direct channel: run paper punch cards while you build one. The $750/year is real money, and the punch card teaches your kitchen staff to track repeat behavior in a way that pays off when you eventually migrate to the CRM-direct path. Don’t spend the $200/month on a standalone loyalty platform thinking it’s the bridge — it isn’t. It’s a worse-margin version of the punch card with extra complexity.

If you’re already running POS loyalty: keep it. The break-even is reasonable and the data is real, even if the data is owned by your POS vendor and gets stranded when you switch processors. Just don’t pay the standalone platform on top of it — that’s a double-spend with overlapping returns.

If you only ever read one comparison from this piece, read this one — the two paid paths most operators actually weigh against each other, side by side on the figures already on the table:

That ordering matters more than the model choice, because each step funds the one after it. Layering a paid loyalty platform onto a broken channel just meters spend against a leak. Here is the sequence I run, and the gate that has to clear before the next step earns its monthly fee.

Don Goldstein is a restaurant operator and runs Muntin Digital. The four-model comparison above is from running each variant on the same restaurant between 2024 and 2025.